I remember evaluating Heap (granted, a long time ago, so I'm sure it's evolved a lot since then) and the auto-tracking capabilities and retroactive analysis sounded awesome, but in practice it wasn't as valuable or easy as I thought. I found that in general, knowing the questions you want to answer and structuring your tracking around that to be far more useful, and that actually being able to take mountains of generic behavioral data and making sense of it wasn't that easy and a ton of work in and of itself. E.g. retroactively finding clicks to a single 'signup button' is easy, but a lot harder when that button is on many different pages and you want to group them or segment them in certain ways, or the implementation changes in some places, etc. In the end, it seemed like it was doing engineering work just to parse the data back out so it seemed more sane to structure in a better way ahead of time.

I'm curious what steps have been taken in the past several years to improve on this and how Freshpaint is approaching this. I think it's a really interesting problem and if done well, could be massively valuable. Good luck!